Ligand Modulated Antagonism of PPARγ by Genomic and Non-Genomic Actions of PPARδ

Background Members of the Peroxisome Proliferator Activated Receptor, PPAR, subfamily of nuclear receptors display complex opposing and overlapping functions and a wide range of pharmacological and molecular genetic tools have been used to dissect their specific functions. Non-agonist bound PPARδ has been shown to repress PPAR Response Element, PPRE, signalling and several lines of evidence point to the importance of PPARδ repressive actions in both cardiovascular and cancer biology. Methodology/Principal Findings In this report we have employed transient transfections and luciferase reporter gene technology to study the repressing effects of PPARδ and two derivatives thereof. We demonstrate for the first time that the classical dominant negative deletion of the Activation Function 2, AF2, domain of PPARδ show enhanced repression of PPRE signalling in the presence of a PPARδ agonist. We propose that the mechanism for the phenomenon is increased RXR heterodimerisation and DNA binding upon ligand binding concomitant with transcriptional co-repressor binding. We also demonstrated ligand-dependent dominant negative action of a DNA non-binding derivative of PPARδ on PPARγ1 signalling. This activity was abolished upon over-expression of RXRα suggesting a role for PPAR/cofactor competition in the absence of DNA binding. Conclusions/Significance These findings are important in understanding the wide spectrum of molecular interactions in which PPARδ and PPARγ have opposing biological roles and suggest novel paradigms for the design of different functional classes of nuclear receptor antagonist drugs.


Introduction
The peroxisome proliferator-activated receptors (PPARs) a, d and c belong to the nuclear receptor family of transcriptional regulators. They function as obligate heterodimers with the retinoid X receptors, RXRs, and signal from PPAR response elements (PPREs) upon binding PPAR-and/or RXR agonists. The PPAR ligands consist of naturally occurring fatty acids and fatty acid derivatives as well as a range of synthetic drugs [1,2,3].
PPARa is involved in the control of catabolic fatty acid metabolism such as peroxisomal b-oxidation and mitochondrial band v-oxidation of fatty acids and is most prevalent in metabolically active tissues such as liver. PPARa is activated by the blood lipid lowering fibrate drugs. These acts as peroxisome proliferators in mice and rats but no adverse effects have been detected in human livers [1,4].
PPARc is involved in fatty acid and glucose homeostasis and is required for adipocyte differentiation and for placental development. Activation of PPARc also seems to act anti-inflammatory and to hinder proliferation or cause apoptosis in cancer cells. The insulin sensitizing thiazolidinedione drugs, which are high affinity PPARc agonists, are used to treat type 2 diabetes and experimentally to treat cancer [5].
PPARd is widely expressed and the most prevalent PPAR in several tissues both in the adult organism and during development [6]. It is also the least known in terms of biological function, although recent reports would suggest that it might have a role similar to PPARa in tissues other than liver. PPARd has also been shown to be involved in placental implantation, wound healing, and carcinogenesis [4,7,8,9]. No PPARd ligands are currently used as such in treatment of disease, although studies on human subjects for the use of a PPARd agonist in the treatment of metabolic syndrome have been reported [10,11].
Recently, it was shown that non-liganded PPARd attracts transcriptional co-repressors when bound to DNA more effectively than PPARa and c. Due to its widespread distribution it was suggested that PPARd acts as a PPRE gateway receptor [12,13]. Given the, sometimes conflicting, results on PPARd biology obtained using various pharmacological and molecular genetic tools we set out to study the ligand modulated antagonism of PPARc1 by genomic and non-genomic actions of PPARd. We found in accordance with [13] that non-liganded PPARd represses PPARa and c. In line with this the PPARd derivative PPARdDAF2, lacking helix 12 (or activation function 2, AF2), acts dominant negatively on PPARa, c1 and d signalling. Furthermore, we found that PPARdDAF2 possess ligand enhanced dominant negative activity on PPRE signalling. In contrast to Shi et al. [13] who reported that a non-DNA binding PPARd derivative didn't exert any dominant negative effects, we found that non-DNA bound PPARd ligand-binding domain (LBD) exerts ligand-dependent dominant negative activity on PPARc1 signalling. Since PPARd and c co-exist in a range of tissues and in many cases have opposite biological effects we propose that the phenomena discovered might have important implications for PPAR experimental designs, PPAR biology in general and possibly drug design.

Results and Discussion
Agonist non-bound PPARd is a repressor of PPARc1 dependent PPRE signalling, but not vice versa Due to its widespread tissue distribution and the fact that it interacts more efficiently on DNA with nuclear receptor corepressors than the other PPAR isoforms it was proposed, as well as demonstrated in vitro, that PPARd functions as a PPRE gateway receptor [12,13]. We confirmed this phenomenon for PPARd and c1 signalling using transient transfection of COS-1 cells with plasmids encoding these PPAR isoforms and a promiscuous (transcriptionally transactivated by all three PPAR isoforms, data not shown for PPARa), PPRE luciferase reporter gene construct (pLFABPluc). We found that the presence of unliganded PPARc1 did not affect PPARd signalling ( Figure 1A) whereas unliganded PPARd significantly (P,0.001) repressed the PPARc1 dependent signalling from pLFABPluc ( Figure 1B).

Ligand-enhanced dominant negative action of PPARdDAF2
Helix 12 modifications (both designed and for PPARc, found in human patients as mutations) have been shown to render PPARs dominant negative due to their inability to recruit co-activators while retaining the ability to bind co-repressors [14,15,16]. Given the superior repressing properties of PPARd, modification of helix 12 should render it a relatively effective ligand independent repressor of PPRE signalling. We have previously employed a PPARd derivative lacking the C-terminal 11 amino acid residues, PPARdDAF2, as a tool for studying PPRE signalling [17]. In order to further characterize the properties of this construct we conducted a range of transient transfection experiments. PPARd-DAF2 was found to act in a dominant negative fashion on PPARa, c1 and d signalling (Figure 2A & B, respectively, P,0.001, data not shown for PPARa), thus confirming and extending our previous observations. Upon agonist binding PPARs undergo a conformational change leading to increased RXR heterodimerisation and shedding of transcriptional co-repressors with the subsequent recruitment of transcriptional co-activators [3]. The increased PPAR-RXR heterodimerisation leads to an increased affinity for PPREs [18,19]. This would in the case of PPARdDAF2 lead to increased occupancy of the PPREs concomitant with recruitment of transcriptional co-repressors and thus further reduced PPRE signalling. We thus investigated the effect of a PPARd agonist on the dominant negative properties of PPARdDAF2. Because of the relatively high endogenous PPRE signalling in the COS-1 cells we employed T47D cells grown in RPMI 1640 medium supplemented with 5% dextran charcoal-stripped serum for this experiment. The effect of over-expressing and transactivating PPARd in T47D cells is shown in Figure 2D. We could detect a small but significant (P,0.001) PPARd (CF dependent) activity in cells with no added PPARd expression vector ( Figure 2C). We could also see a small but significant (P,0.01) effect of introducing PPARdDAF2 on non-CF dependent transcription of the luciferase gene in pLFABPluc ( Figure 2C). The dominant negative effect of introducing PPARdDAF2 into the system was further enhanced by the addition of CF (P,0.001). This indicates that for PPARdDAF2 CF acts as an inverse agonist that enhances the dominant negative effect, a novel concept for type II nuclear receptors. The concept was discussed and investigated for the only PPARd antagonist described to date, GSK0660. GSK0660 did not, however, increase occupancy of PPARd or transcriptional corepressors to chromatin PPREs [20].
The PPARd ligand-binding domain is a repressor of PPRE dependent PPARc1 signalling in the presence of a PPARd agonist Since the PPARs act as RXR heterodimers it would be conceivable that RXR competition could occur among the PPAR isoforms. In fact, ligand dependent RXR competition has been described for PPARa and liver X receptor (LXR) [21,22], PPARb/d and LXRa [23], PPARa and thyroid hormone receptor (TR) [24] as well as PPARc and TRa1 and b mutants [25,26]. Agonist-bound wild-type PPARd and c activate transcription when bound to PPREs. Thus, in order to study the PPRE independent effects of PPARd and c we needed a non-DNA binding derivative with a functional ligand binding and activating domain. We generated an expression plasmid for the PPARd LBD, pCLDN-dLBD, and tested it for the desired properties in a mammalian two-hybrid assay. Co-expression of the GAL4-RXRa fusion protein and the PPARd LBD led to CF induced upstream activating sequence (UAS) dependent transcriptional transactivation, strongly indicating that the PPARd LBD is functional with respect to RXR heterodimerisation and transcriptional coactivator recruitment ( Figure 3A, P,0.001).
Subsequent to the functional validation of the PPARd LBD we investigated whether it had a dominant negative effect on PPARd and c1 signalling. We found that PPARd but not PPARc1 signalling was abolished by co-expression of the PPARd LBD ( Figures 3B (P,0.001) and C, respectively). One important Given the known effects of agonist binding to a PPAR one could speculate whether the dominant negative effect of the PPARd LBD is due to RXR or transcriptional co-activator squelching. To address this question we co-expressed RXRa and the transcriptional co-activator, steroid receptor co-activator 1a (SRC1a), with PPARd and c1 with and without the PPARd LBD. PPARd signalling was found to be repressed by co-expression of the PPARd LBD ( Figure 4A and B, P,0.001 and P,0.05, respectively). This dominant negative effect was abolished by co-expression of RXRa ( Figure 4A, P.0.05). Co-expression of SRC1a with PPARd increased the agonist dependent inducibility of reporter activity but didn't abolish the effects of PPARd LBD dependent repression ( Figure 4B).
We then proceeded to study the effect of RXRa and SRC1a coexpression on the effect of the PPARd LBD on PPARc1 signalling. In this experimental setup the PPARd LBD showed dominant negative behaviour in the absence of CF (Figures 4C and D, P,0.001 and P,0.05, respectively). The dominant negative effect of the PPARd LBD was somewhat enhanced by the PPARd agonist ( Figures 4C and D). The effect of co-expression of RXRa was similar to that of the PPARd experiment with overall activity somewhat increased but with lower levels of PPARc agonist dependent induction and in abolishing the dominant negative effect of the PPARd LBD ( Figure 4C). Co-expression of SRC1a increased the level of activity of PPARc1 without having a much of an effect on the level of induction ( Figure 4D). The PPARd LBD repressed PPARc1 signalling (P,0.05) with additional repression seen in the presence of CF ( Figure 4D). As was the case for PPARd, the addition of SRC1a increased the overall levels of signalling ( Figure 4D). Also similarly with the SRC1a coexpression experiment with PPARd the addition of SRC1a did not abolish the PPARd LBD mediated repression. Instead, the level of PPARd LBD mediated repression became more pronounced ( Figure 4D, P,0.001). Furthermore, the PPARd agonist enhanced repression was more marked (Figure 4D, P,0.05). Since the addition of RXRa seems to relieve the PPARd LBD mediated repression of PPARd and PPARc1 signalling whereas the addition of SRC1a still allows the PPARd LBD mediated repression we conclude that RXR sequestration is likely to be the main mechanism behind the phenomenon. We thus speculate that ligand dependent RXR competition could occur in vivo between at least PPARd and PPARc and quite possible between all three PPAR isoforms.

Concluding remarks
The major conclusion we draw from this study is that care must be taken when interpreting results obtained from all genetic models of PPARd action. The genetic ablation of PPARd will remove both the ability to activate PPARd, but also the intrinsic role that PPARd has in the tempering of PPARa and PPARc signalling. Therefore it is prudent to use a wide range of both gain and loss of function experiments in order to fully understand the function of PPARd and its relationship to PPARa and PPARc signalling. This is most likely to be true for other nuclear receptors forming heterodimers with RXRs as well.
Our study also might suggest a novel paradigm for the design of different functional classes of type II nuclear receptor antagonist drugs. One could envisage two sets of nuclear receptor antagonists with very different biological actions (simplistically stating the two extremes of antagonist behaviour); one that displaces the PPAR/ RXR complex from the PPRE and one that simultaneously increases DNA binding and transcriptional co-repressor recruitment.

Cloning and plasmids
General DNA techniques were performed according to [27]. DNA sequencing was done by the DNA Analysis Facility, Human Genetics Unit, at Ninewells Hospital, Dundee. Escherichia coli XL1 Blue was transformed according to the manufacturer's instructions (Stratagene).

Growth of cells and transient transfections
COS-1 and T47D cells (Cancer Research U. K. cell resources unit) were grown in a 5% CO 2 atmosphere at 37uC in high glucose DMEM supplemented with 10% foetal bovine serum and 50 U/ml penicillin G and 50 mg/ml streptomycin (Gibco) and 2 mM L-glutamine for COS-1 and T47D cells, respectively. For transfections the T47D cells were grown in RPMI 1640 (phenol red-free) containing 5% dextran-charcoal stripped foetal bovine serum. Transient transfections of COS-1 cells and T47D cells were performed in six-well plates using DEAE-dextran according to Cullen [35] and Lipofectamine 2000 (Invitrogen), respectively. 24 hours post transfection, medium containing 50 nM compound F, CF, [33] for PPARd activation and/or 500 nM rosiglitazone, BRL, [36] for PPARc1 activation in a final concentration of 0.1% dimethyl sulfoxide (DMSO) or DMSO alone was added. 48 hours post transfection cell lysates were generated using Promega's reporter lysis buffer.
For all transfections 500 ng luciferase reporter (pLFABPluc or p46UAS-TK-luc) and 50 ng pSVb-galactosidase were used per well in six-well plates. Luciferase activity was assayed with the Promega luciferase assay substrate and b-galactosidase activity according to Sambrook et al. using o-nitrophenyl-b-D-galactopyranoside [27] or using the chemiluminescent b-gal reporter gene assay kit from Roche.

Statistical analysis
Relative reporter gene expression is stated as the luciferase activity normalized against the corresponding b-galactosidase activity. These values have in turn been normalised against the mean of the normalized luciferase activities of the leftmost bars in each graph. Each experiment was repeated three times and the bars in the graphs represent the means and the error bars represent the standard error of the mean. One-way ANOVA was performed on the data from each experiment and the Newman-Keuls test was employed for calculating statistical significance using GraphPad Prism 3 software.